Show HN: Bibfixer – AI-powered BibTeX cleaner (github.com)

Bibfixer is a new Python CLI tool (pip install bibfixer) that uses LLMs plus web search to clean, complete and standardize BibTeX files. Point it at an input .bib (bibfixer -i input.bib), optionally set an output file (-o) and style preferences (-p) or supply a custom prompt (--prompt-file). It can fill missing metadata (authors, venues, page ranges), normalize title casing and acronyms (e.g., "llm" → "LLM"), and convert arXiv entries into published inproceedings with booktitle, volume and page details. Prompts/default.md is editable so you can enforce project conventions (e.g., “Use NeurIPS instead of NIPS”). For the AI/ML community this addresses a persistent pain point: messy Scholar-derived BibTeX and inconsistent citation styles across papers, repos and supplements. By automating lookups and style rules, Bibfixer saves time and improves readability and reproducibility of bibliographies used in papers, code releases and reviews. Technically, it’s a light CLI wrapper around LLM + search results with customizable revision instructions; it requires an OpenAI API key (export OPENAI_API_KEY=...) and includes a warning that automated metadata can be incomplete or incorrect. Users should review outputs before submission and can quickly diff inputs and outputs (diff -y --suppress-common-lines input.bib output.bib) to spot changes.